Adding content to a wskill

§ 1Purpose

The core capture loop: decompose a piece of knowledge, classify it, write the unit, link it, pin it, render.

§ 2Prerequisites

• An existing wskill folder that checks clean (just wskill-check).

§ 3Flowchart

§ 4Steps

Break the knowledge into single-idea pieces — one idea, thing, value or task per note. If a draft note needs an "and" in its summary, split it. See Decomposing information.

Run each note through the decision guide: a dated finding from an investigation → research (see Capturing research); a repeatable task → procedure; a concrete NAMED thing (person, software, place, …) → entity with a kind from schema/kinds.wcl; an indisputable value/table → fact; a word needing a definition → term; otherwise → concept. Never default to entity.

// data/reference/<id>.wcl (or the per-kind folder)
concept <id> {
  name     = "<Headline>"
  summary  = "<One-liner the indexes show.>"
  audience = :both              // opt into the skill when the agent needs it
  related  = [<other_ids>]
  body { p "The substance — capture it here, never defer to an external source." }
}

Write the block instance in data/ (one file per unit as the wskill grows; add the import line to the folder's main.wcl). Give it a stable id, a headline, a summary, and a self-contained body. Attach worked examples as example blocks with unit = <id>.

Fill related with the ids of the units this one leans on or explains — links resolve both ways (each page also lists what references it). A unit with no links is usually either misfiled or not atomic.

Add the id to the related list of the index that owns its area (or create a new index branch — see Building the wskill index). Unpinned units stay reachable via links but invisible in the nav.

$ just wskill-check && just render

wcl check catches schema violations (a wrong entity kind, a missing required field) with file/line errors; the render makes the new pages. Fix anything it reports before moving to the next note.